Cotton yield and its market value are of prime concern to producers. Price fluctuation in time and space are governed by the world economy, environmental conditions and management practices. These parameters have a direct impact on farmer’s investment decision for a short and long-term planning. The immediate concern for growers is the crop yield. The analysis of spatial and temporal yield variability is fundamentally important for growers and. The objective of this investigation is to describe the temporal yield variability of cotton in Arizona. Cotton yield in Arizona counties from 1935-1999 was used for semivariogram and fractal analysis. The semivariogram was calculated for each county for Pima and Upland cotton for different year interval (h). The slope of log(semivariogram vs log h was used to calculate fractal dimension D, an indication of variability pattern. D close to 1 indicates the likelihood of a long- term variation. A D close to 2 indicates a short-term variability. From 1935 to 1990 the cotton yield has increased two to three folds in every county. The increase in yield is the result of plant breeding, better plant and soil nutrient management, herbicide, insecticide use and irrigation technology. The increase in yield also increased the risk of temporal and spatial fluctuation. Fractal dimension increased from 1.473 in Maricopa to 1.794 in Graham County for Upland cotton. The Fractal dimension increased from 1.457 in Yuma County to 1.781 in Cochise County for Pima cotton. Small D value in Yuma County is indicative of long-term yield variability while higher D value in Graham County, indicates short-term variability. Fractal analysis appears to be a useful tool to quantify temporal variability. It has a potential application for agronomic research and production in assessing variability in soil and plant parameters.